Skip to main content

CPU vs GPU

Overview​

The Hiya Voice Verification engine ships in two variants: a CPU image and a GPU image. Both expose the same API and produce the same results — the difference is the compute target they are optimized for. Choose the variant that matches your infrastructure.

When to choose CPU​

The CPU image is a good fit when:

  • You are running small or medium deployments where GPU infrastructure is not already in place.
  • You prefer simpler infrastructure with no GPU driver or GPU operator dependency.
  • You are deploying on general-purpose cloud instances or on-premises servers without discrete accelerators.
  • You want to minimize operational complexity during initial rollout.

Recommended hardware: Intel Emerald Rapids (5th-generation Xeon Scalable). Other modern x86_64 architectures are supported — see Installation Requirements or contact Hiya for guidance on your specific hardware.

When to choose GPU​

The GPU image is a good fit when:

  • You are running larger deployments where GPU infrastructure is already available or planned.
  • You want a more cost-efficient option at scale, where GPU acceleration reduces the number of instances needed.
  • You are targeting production deployments at scale and want the recommended configuration.

Recommended hardware: NVIDIA L4. Other GPUs may work but have not been validated at the same level; contact Hiya for guidance on alternative GPU architectures.

Compute targetRecommended hardwareNotes
CPUIntel Emerald Rapids (5th-gen Xeon Scalable)Other modern x86_64 architectures supported; contact Hiya
GPUNVIDIA L4Other GPUs may work; contact Hiya for alternative GPU guidance

Next steps​

Once you have chosen a compute target, pull the corresponding image (see Container Image) and follow the deployment guide for your platform (see Deployment Overview).

Contact your Hiya account team to discuss workload-specific sizing and performance characteristics.